Яндекс.Метрика

П.А. Дергач, Г.Н. Логинов, С.В. Яскевич, Н.А. Ульянов,Е.Е.Пятаев

Издание: EAGE. Геомодель 2021: 23-я конференция по вопросам геологоразведки и разработки месторождений нефти и газа (г. Геленджик, 6-10 сентября 2021 г.): Тезисы докладов
Место издания: Геленджик , Год издания: 2021
Страницы: 1-5

Аннотация

In this paper, the convolutional neural network algorithm for the detection of the signals from weak local earthquakes is presented. Traditionally, algorithms based on the classical STA/LTA methods are used for the detection of the signals from earthquakes. Recently, algorithms based on neural networks were used too. However, existing solutions are designed to detect and determine the arrival times of waves from strong earthquakes and are ineffective at a low signal-to-noise ratio. The work presents the detailed structure of the neural network, as well as the processes of preprocessing and training the network. It is important that the datasets for the training and for testing the algorithm were taken from different local networks of seismological monitoring. A comparison of the results of testing the algorithm on real records was carried out with the MER algorithm. It is shown that, in contrast to MER, on continuous records with a duration of three days, the proposed algorithm worked without false triggers.
индекс в базе ИАЦ: 035529